Voter turnout in contested races is the real statistic
It's always reported how low US voter turnout is in midterm elections. 2006, at about 40%, seems pretty poor, though it was higher than 2002.
However the statistic I would like to see is "Voter turnout in districts where there is an important, hotly contested race." This is the number we might want to monitor from year to year.
Virginia, it turns out, which had the Webb-Allen "Macaca" race, had the highest voter turnout in its history. You wouldn't think that after hearing about the low turnout of a typical mid-term. Of course it will also go down as the first time a major U.S. politician was taken down due to blogs, the web and YouTube. Since it was so close, almost any factor can be given credit for Allen's loss.
It is not surprising that when there is no contested race, that turnout is low. The U.S. for various bizarre reasons, has most incumbents always safe in their seats. This switch of 30 or so seats in the house and 6 in the senate is considered a major upheaval, nigh a revolution, by Americans. With seats so safe, there is no suprise there is little incentive in voting. U.S. ballots are very complex compared to many countries, and there are often long voting lines, and you don't get official time off to vote.
Contrast that to Canada, where a public upset with the Conservative party's introduction of the visible Goods and Services Tax (a 7% VAT) took the party from having a majority of parliament to having TWO seats. 2, as in 1 plus 1. There's no such safety zone for incumbents, no cry for term limits in much of the rest of the world. There, if the public gets upset it throws the bums out, or drops them back to a minority position due to the fact that there are more than 2 parties.
I hope one of the major statistical agencies starts tracking voter turnout modulated by how motivated the voters are in particular districts. Of course voter turnout is the final metric of how motivated they were, but there are other, earlier indicators in most cases.